Electronically processing image data often encounters a variety of problems during Optical Character Recognition (OCR). Many times the OCR engine can identify a set of pixels as a particular character or digit with some level of confidence by without complete confidence.
For decent sized scan jobs of several documents, a single error in an accurate recognition can create a sizable manual task in locating the problem document where the recognition was inaccurate.